Fibrillogenesis, or the assembly of the extracellular matrix (ECM), involves both active cell-mediated polymerization processes as well as cell-mediated remodeling processes (McKeown-Longo and Mosher,1983; Wierzbicka-Patynowski and Schwarzbauer,2003). Protein-protein interactions that mediate the polymerization of fibrils from individual proteins or protein oligomers have been subject to intensive biochemical analysis (Sechler et al.,2001; Takahashi et al.,2007). Additionally, the earliest steps of fibril formation have been elucidated in controlled cell culture systems (Ohashi et al.,1999,2002). With few exceptions, histological approaches have been used to study the macroscopic processes of ECM assembly during embryonic development, wound healing, organogenesis, and oncogenesis.
On the molecular level, multimerization of two fibronectin dimers requires physical stretching of at least one of the interacting dimers (Halliday and Tomasek,1995; Baneyx et al.,2002; Krammer et al.,2002). While sites have been identified in fibronectin that mediate this interaction, a precise molecular mechanism is still lacking. The physical properties of fibronectin molecular subunits have been studied with both molecular dynamics simulations as well as biophysical observations of single molecules under tension (Craig et al.,2001; Gao et al.,2002; Krammer et al.,2002; Oberhauser et al.,2002; Abu-Lail et al.,2006). Thus, the principles of fibril assembly involve both specific protein–protein interactions as well as the physical application of force and strain to fibril networks. The magnitude of the forces required for polymerization exceed those available from thermal or chemical processes (Baneyx and Vogel,1999).
In general, the large forces required for fibronection oligimerization can only be generated by active cellular processes linked to the cytoskeleton. Attachment of cells to the ECM is largely accomplished through the integrin family of receptors (for review, see Dzamba et al.,2002). ECM-specific integrin heterodimers link the cytoskeleton within the cell to the ECM. These linkages can control both the timing and location of cell–ECM interactions by integrating the cytoskeleton to both intracellular and extracellular signaling pathways. Once a mechanical linkage is formed, cells may generate force on the ECM through several different mechanisms: through the application of traction forces commonly associated with myosin II-based contraction found at the rear of lamelliform protrusions (Horwitz and Parsons,1999), through transport processes associated with retrograde flow of subcortical actin meshworks (Suter et al.,1998), or through the directed transport of cargo and membrane associated proteins by microtubule- and actin-based motors such as dynein, kinesin, or various myosins (Sheetz,1999). Thus, fibril assembly requires the integration of cell-ECM adhesion and cell-force generation with the biophysical and biochemical processes of ECM components.
Uncontrolled fibril assembly is prevented through a balance between cell traction forces and the physical-mechanical properties of the ECM (Harris,1987). The early steps of matrix assembly proceed as a cell protrudes into the network and samples the stiffness of local fibronectin fibrils with focal attachments. Fibrils that are oriented along the direction of traction and anchored by other attachments resist the traction generated by the attachment and strengthen the contact (Riveline et al.,2001). Forces applied through the attachment by the cell then pull and stretch the fibril. Strain in the fibril, rather than stress, controls the rate of polymerization (Baneyx et al.,2002; Krammer et al.,2002). Cells may then either release the fibril or remain attached to the fibril after maturation (Zamir et al.,2000). Continuing rounds of cell protrusion, traction, fibril strain, and polymerization add to fibril thickness and increases its mechanical stiffness. Forces generated by cell traction may increase during these rounds as the thicker, stiffer fibrils attract and strengthen nascent attachments (Sheetz et al.,1998; Galbraith et al.,2002). However, at some point the forces generated by the cell are no longer able to stretch the fibril to a degree required for additional polymerization. Thus, the coupling of cell protrusions, formation of attachments to the ECM, cell traction, fibril strain, and strain-dependent polymerization form a feedback control system for the homeostatic maintenance of the ECM.
The integration of cellular, biophysical, and biochemical processes that modify the ECM dominate the macroscopic processes of oncogenesis, wound healing, organogenesis, and morphogenesis. Of these macroscopic processes, morphogenesis in amphibians involves some of the most dynamic changes in the ECM. In a matter of hours, entire fibrillar networks are laid out and remodeled (Nakatsuji et al.,1985; Winklbauer and Keller,1996; Darribere and Schwarzbauer,2000). Sequential assembly of ECM defines tissue boundaries, establishes the embryonic body plan (Lee et al.,1984; Wedlich et al.,1989; Marsden and DeSimone,2003; Davidson et al.,2004b), and migratory tracks for neural crest and growth cones (Epperlein et al.,1988; Krotoski and Bronner-Fraser,1990). Furthermore, dynamically assembled ECM plays a key role during branching morphogenesis in the salivary gland (Sakai et al.,2003) and is likely to contribute to shaping incipient organs such as the heart, liver, vasculature, and kidneys.
In this study, we present visualization techniques that allow the direct observation of both cells and the extracellular matrix as fibronectin fibrils form in embryonic tissue explants. We characterize several macroscopic processes that remodel the fibrillar fibronectin matrix with different timescales and demonstrate the role of active cellular processes that maintain and localize the fibronectin matrix to specific tissue interfaces.
Embryonic Tissue Explants Develop Complex Fibrils
To visualize fibril dynamics within a live tissue, we modified previous explant preparations and visualization techniques (DeSimone et al.,2007). Histological analysis had identified a phase of rapid assembly and remodeling of a fibronectin-rich fibrillar ECM during early stages of gastrulation in Xenopus laevis (Davidson et al.,2004b). To visualize fibril dynamics at these stages, we turned to using well-characterized explants of the frog embryo's animal cap or marginal zone (Fig. 1A; Davidson et al.,2004a). Animal cap explants contain prospective ectoderm whereas marginal zone explants contain prospective ectoderm and mesoderm. In vivo both ectoderm and mesoderm appose a fibrillar fibronectin matrix. When cultured on artificial rigid substrates (e.g., plastic, glass, or fibronectin-coated glass/plastic) these explants continue to secrete fibronectin (FN) which adsorbs onto these nondeformable substrates but do not develop a fibrillar ECM (see Supplementary Figure S1, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat). To visualize matrix remodeling under conditions more representative of deformable substrates in vivo, we cultured either marginal zone or animal cap explants between sheets of agarose (Fig. 1B) supplemented with 0.1 mg/ml of FN in the medium. Tissue explants do not adhere to agarose sheets and develop a complex fibrillar ECM. A key difference between these explants and previous preparations is that we have removed a single layer of epithelial cells to prevent a wound healing response and allow direct observation of forming fibrils using time-lapse laser scanning confocal microscopy (see also Davidson and Wallingford,2005). If the epithelial layer of cells is not removed wound healing begins and the epithelial sheet engulfs exposed deep cells rapidly blocking the view of forming fibrils. The fibrillar ECM that forms under these conditions is confined to the surface of the explant. With the exception of the early to mid-gastrula stage animal cap ectoderm that assemble fibrils along a surface that faces the fluid-filled blastocoel, most ECM is assembled at interfaces between two tissues. Fibril assembly on exposed mesodermal tissues in explants takes place without either the endoderm or ectoderm tissues that would normally face the mesoderm at these interfaces.
Fibrils Form and Remodel in Close Association With Underlying Cells
To visualize both cells and fibronectin fibrils, we expressed mRNA encoding a plasma membrane localizing enhanced green fluorescent protein (EGFP; GAP43-GFP) in embryos and cultured explants in media containing a Cy3-conjugated monoclonal antibody directed against frog fibronectin (Cy3-4H2). Fibronectin fibrils on mesodermal cells in time-lapse confocal sequences appear as an interconnected meshwork that forms between the cellular tissue and the agarose sheet (Fig. 1D). Like the assembly of fibrils in vivo, fibrils rarely penetrate between cells (see supplemental Fig. S1). Both lamellipodia and filopodia can be tracked in scattered-labeled individual mesodermal cells (Fig. 1E; scattered-labeled cells are surrounded by unlabeled cells). Both cells and the fibrillar matrix assembled at their surfaces move dynamically in this sequence (see Supplementary time-lapse Movie S1A). Because these time-lapse sequences are collected with a confocal microscope small fibers occasionally “disappear” but are in fact simply moving out of the confocal microscope's plane-of -focus. Fibrils can transiently disappear because the uneven surface of the explant or the agarose gel provide spaces for fibrils to occasionally escape the plane of focus. Similar fibronectin fibril dynamics are seen at the surface of ectodermal tissues (see Supplementary time-lapse Movie S1B).
Growth and Topological Elaboration of Fibrillar Network
Time-lapse sequences of fibronectin fibrils on both ectoderm and mesoderm reveal that the network undergoes distinct topological changes and that each type of change occurs with its own characteristic timescale. The most obvious change in the network is growth (Fig. 2A; see Supplementary time-lapse Movie S2). Over the course of hours the fibrillar matrix grows in density and complexity. Most fibril growth may reflect polymerization rather than simple “thickening” after fibrils are released and shorten in length (see fibril “loop” between arrowheads in Fig. 2B; see supplementary time-lapse movie S3). Both polymerization and shortening-and-thickening produce a denser fibrillar array, however, they do so on different timescales. Polymerization takes hours to increase the density of fibrils while individual fibrils can shorten and thicken in as little as 10 min (compare timescales in Fig. 2A,B). The “connectivity” of the fibrillar network, that is, the number and location of connections between fibrils, changes on the timescale of minutes. Single fibrils can move, contact, and anneal to neighboring fibrils during this time (arrow in Fig. 2C; see Supplementary time-lapse Movie S4). These new connections then appear to integrate immediately into the mechanical fabric of the matrix (e.g., new fibril links stretch by last frame in Fig. 2C). The most rapid change in the network occurs in seconds as fibrils stretch and break (arrow indicates intact fibril in Fig. 2D; see Supplementary time-lapse Movie S5). The free ends of these broken segments then recoil into globular particles (arrowheads after fibril segment breaks in Fig. 2D) perhaps through a process of self-annealing. Thus, fibronectin fibrils form a highly dynamic network that is undergoing constant remodeling by a series of topological processes in which the fibrils polymerize, anneal, stretch, and break.
Elaboration of the Network Is Driven Through Cell Contact With Fibrils
Movement of the fibrillar network occasionally correlates with cell protrusions. Lamellipodia in these explant preparations extend from one cell under the ventral surface of neighboring cells. The majority of lamellipodial movements occur without apparent contact with fibrils; however, some appear to contact, bind, and drag fibrils over the course of a protrusive duty cycle (Supplementary time-lapse Movie S6A). Several fibril segments (within yellow circles; Fig. 3A,C) move from the surface of one cell (hash mark in Fig. 3B) after contact with a lamellipodia (arrowhead at 6 min in Fig. 3B) and move to the boundary between two cells (between the asterisk and hash marked cells in Fig. 3B) as the lamellipodia retracts. Fibrils that run along cell–cell boundaries are subject to additional modification as small fibril “spines” frequently form perpendicular to the main fibril (see arrowheads in Fig. 1B). Additional fibril movements coincide with filopodia (data not shown), however, many fibril movements are not associated with either lamellipodia or filopodia.
In most cases, fibril movements cannot be correlated with cell protrusions but instead move toward focal points far from cell–cell boundaries (asterisk; Fig. 3B; see Supplementary time-lapse Movie S7; for additional examples, see fibril movements in Supplementary time-lapse Movies S2A, and S2B). These fibrils are gathered to a central, perinuclear location on the ventral cell surface (asterisk in Fig. 3E, or within yellow circles in Fig. 3D,F). Fibril gathering focal points are relatively static structures; however, fibrils can join or leave them on the timescale of minutes (e.g., see fibril movements at the end of supplementary time-lapse movie S2). Movements of the fibrillar network, distinct from cellular protrusions, continue even after a dense fibrillar network is assembled (see last 10 min of elapsed time in supplementary time-lapse movie S1).
Maintenance of the Network Depends on F-actin
Because the fibrillar network was so closely associated with the cell surface, we wondered whether the cortical actin cytoskeleton was essential to maintain the matrix. Cellular protrusions, as well as other force generating machinery in the cell, require polymerization of the actin cytoskeleton. We inhibited actin polymerization by adding cytochalasin D (CD) to explants that had already assembled a fibrillar network. Twenty minutes after addition of 4 μM CD to cultured explants the fibrillar network collapses to large thick cables, cells in the explant become round and cease protrusive activity but remain attached to one another (Fig. 4A; see supplementary time-lapse movie S8). These cables remain associated with the exposed cell surface of the explant and are not found deeper within the explant even after cells round up and cell–cell contacts are significantly reduced. When intact fibril networks are incubated in 1 mM of the divalent cation chelator ethylenediaminetetraacetic acid (EDTA), the fibrils also collapse to cables, however, these cables are no longer held to the outer surface of the explant, penetrate between cells, and move into the middle of the explant (Fig. 4B; see Supplementary time-lapse Movie S9). Thus, the F-actin cytoskeleton assists in maintaining the open network topology but is not required to hold the fibrils to the outer surface of the explant.
Multiple Steps for Fibril Assembly and Remodeling May Form the Basis of Homeostatic ECM Maintenance
Distinct processes of growth, annealing, stretching, breakage, and elastic recoil shape the ECM that forms on the exposed surface of frog embryonic tissue explants. These observations support a three-step process for the formation of the fibrillar matrix (Fig. 5A) in which FN-puncta are stretched both across apposing cell–cell boundaries or within the membrane of single cells (Dzamba et al., submitted). These short fibrils can then anneal to one another to form dense contiguous networks that span multiple cells. The incipient network is then remodeled as fibrils are both gathered toward the cell's perinuclear region on the ventral surface of cells and as fibrils are swept to cell-cell junctions (Fig. 5B). Thickening of individual fibrils may occur through both polymerization or as released fibrils shorten and thicken (Fig. 5C). Further remodeling of the network can occur through cycles of annealing–breaking–recoiling (Fig. 5D). This study is just the first step in describing these apparently chaotic topological changes in the ECM and future studies will be needed to determine the precise role of cell-adhesion and cell-traction in both polymerizing and regulating fibrillar ECM assembly.
The physical coupling of ECM and cells' actin cytoskeleton that underlies the mechanics, assembly, and remodeling of the ECM may be responsible for the homeostasis of the ECM in both developing embryos and in adult tissues. The concept of ECM homeostasis was first suggested by Harris (1987) to explain the capacity of sponge tissues to maintain an unchanging ECM, despite the processes of growth and disturbance. In the current study, we report live changes in the fibronectin-ECM as the matrix assembles and remodels in frog embryonic tissue explants. Earlier studies related the effects of molecule-scale strain and force in exposing cryptic sites within FN required for fibril assembly (Schwarzbauer and Sechler,1999). Thus, our findings support a direct role for cell attachment in assembling and remodeling the fibrillar FN-ECM in tissues as part of a strain-dependent feedback mechanism for the regulation of ECM density and topology (McKeown-Longo and Mosher,1983; Schwarzbauer and Sechler,1999).
Our time-lapse sequences of cell and ECM dynamics during morphogenesis should dispel the perception that the extracellular matrix is static during morphogenesis. Recent work has shown that fibronectin fibrils are extensively remodeled during gastrulation in the frog (Davidson et al.,2004b) and chick embryos (Czirok et al.,2004; Zamir et al.,2006) even as cells undergo migration and rearrangement. The timescale of matrix remodeling is nearly the same as the timescale of the cell rearrangements that drive convergence and extension (Keller et al.,2000). Fibronectin fibrils may be remodeled on similar timescales during bone remodeling by osteoblasts (Dallas et al.,2006; Sivakumar et al.,2006) and during branching morphogenesis in the submandibular salivary gland (Sakai et al.,2003; Larsen et al.,2006). In all these examples, cells may migrate on, remodel, and move ECM with them. Potentially, information embedded in the ECM such as bound growth factors will be likewise moved or remodeled with consequences for cell signaling and behaviors. Future advances in image processing, colocalization of focal adhesion proteins, and biophysical analyses of the mechanics of the ECM will be needed to understand the cell biology underlying the complex topological changes in the ECM and their role in developing tissues. The interplay between fibrils, which may mechanically couple adjacent cells, and the protrusive activity that drives these cell rearrangements is not well understood but can now be clearly resolved.
Frogs, Eggs, and Explant Preparation
Fertilized frog eggs were obtained using standard methods, dejellied, and cultured in 1/3× Modified Barth's Solution (Sive et al.,2000) to the desired stage (Nieuwkoop and Faber,1967). At early gastrula stages, embryos were selected and transfered to explant culture medium (Danilchik's For Amy, DFA; Sater et al.,1993) for microsurgery, then cultured for longer periods in DFA.
Explants consisting of deep ectoderm or mesoderm were isolated using hairloops and eyebrow knives. Embryos were manually removed from their vitelline with forceps. Marginal zone explants are prepared as described (Davidson et al.,2004a) with slight modification. The outer, single cell-layered epithelium is removed once the mesendoderm is pulled away from the interface that forms the Cleft of Brachet (see Fig. 1B in Davidson et al.,2004a). The explant is then separated from the remainder of the embryo and transfered to confocal viewing chambers containing DFA supplemented with 0.1 mg/ml bovine plasma fibronectin (Roche Molecular Biochemicals), antibotic/antimycotic (Sigma), and 5 μg/ml Cy3-4H2 (see antibody preparation).
Once cut, explants were cleaned of debris and mounted in chambers for confocal imaging. Chambers were made from rectangular-shaped custom pieces of acrylic milled to provide small volume for culture media. Acrylic chambers were glued to large coverslip fragments using silicone grease (Dow-Corning) and sealed with small square glass coverslips. For experiments where media needed to be added a small corner of the top coverslip was removed leaving a small opening. Explants were transfered to the well along with two thin rectangular sheets of agarose. The explant was then sandwiched between the agarose and held in place by a small coverslip fragment and silicone grease.
Expression of Membrane-Localized GFP
Synthetic capped RNA encoding a plasma-membrane localizing GFP (GAP43-GFP) was transcribed (Epicentre) from plasmid (Moriyoshi et al.,1996). To obtain embryos expressing GAP43-GFP evenly throughout the explant, embryos were injected 4 times around the marginal zone with a total of 0.2 ng of RNA before the first cell division after fertilization. To obtain scatter-labeled samples, embryos at early cleavage stages of 16 to 128 cells were injected in single blastomeres in the marginal zone with the same equivalent concentration of RNA.
Antibody Isolation and Conjugation
For live imaging of fibronectin fibrils, the mAb 4H2 (a nonfunction blocking antibody raised against the central cell binding domain of frog fibronectin (Ramos and DeSimone,1996) was conjugated to the Cy3 fluorophore using an NHS-ester form of Cy3 (FluoroLink; Amersham Pharmacia Biotech). Cy3-4H2 was dialyzed against DFA (without BSA) and concentrated with cutoff filters (Centricom YM-30 centrifuge filters; Amicon).
Imaging and Image Processing
Confocal time-lapse sequences were collected using a confocal laser scan head (PCM2000, Nikon, Melville, NY) mounted on an inverted compound microscope (Nikon). High resolution images were acquired with a 1.40 n.a. ×60 oil-immersion plan apochromat objective. The GFP signal was excited using the 488 nm line from an argon laser and collected with the 515-nm emission filter. The Cy3 signal was excited using a 543-nm Green HeNe laser and collected with the 565-nm emission filter. Independent neutral density filters were used to reduce the excitation intensities to minimal levels to reduce photodamage to the samples. Furthermore, the confocal pinhole was opened to the “middle” position to collect from a larger z-slab and to increase brightness. Samples were exposed to both excitation wavelengths and the emission signals collected from the same scan. Multi-wavelength, multi-Z-position, and time-lapse sequences were collected and stored as uncompressed animation stacks on a computer (PC) running image acquisition software (Compix Inc., Cranberry Township, PA). These stacks were transfered to a computer (PC) and processed using ImageJ (ImageJ; Wayne Rasband, NIMH; see http://rsb.info.nih.gov/ij/). Processing was limited to linear adjustments to the lookup table and to one round of smoothening and one round of sharpening.
We thank Charlie Little for sharing the technique for live-labeling ECM. We would also like to thank Eddy DeRobertis and Jeffery Miller for providing plasmid for GAP43-GFP. Confocal facilities were provided by the W.M. Keck Center for Cellular Imaging at the University of Virginia. D.W.D., R.E.K., and L.A.D. were funded by The National Institutes of Health.